An Algorithm for Discretization of Real Value Attributes Based on Interval Similarity

Table 3

Comparison of C4.5 performance on 9 UCI real datasets.

Datasets	Predictive accuracy (%)			Number of nodes			Number of rules
	EXT	SIM	Boolean	EXT	SIM	Boolean	EXT	SIM	Boolean
Iris	91.67	93.67	90.0	20.85	20	20.43	14.35	13.7	13.03
Glass	51.16	55.12	49.74	121.5	125.6	127.8	79.55	112.4	105.6
Breast	92.55	94.12	92.0	89.1	79.1	88.3	57.5	45.1	58.2
Wine	80.28	91.53	88.9	62.8	28.65	48.35	36	21.4	28.56
Auto	77.15	78.73	72.2	139.65	122.9	148.6	99.35	91	104.5
Bupa	45.29	38.41	38.24	236.4	248.7	249.3	183.55	201.9	196.9
Machine	77.38	83.69	66.7	64.15	62	72.35	42.45	42.9	48.36
Pima	61.82	59.25	63.6	448.4	485.9	445.3	342.1	427.6	356.8
Ionosphere	83.94	88.48	85.9	90.25	74.3	82.23	1	1	1